Motor Assessment after Stroke

The project focuses on developing and validating video-based pose estimation tools that use artificial intelligence to automatically track body movements for neurologic motor assessments. Pose estimation identifies keypoints of the body—such as elbows, knees, and fingers—from ordinary digital video, enabling quantitative analysis of movement without the need for costly motion capture systems or limited wearable devices. The goal is to create accessible tools that can provide full-body motor assessments in clinical and remote settings using only smartphones or tablets.The team employs OpenPose, a state-of-the-art, open-source algorithm, to track 25 body keypoints and 21 hand keypoints. Custom workflows convert video keypoint data into measurable outcomes, allowing comparison against motion capture systems and clinical standards. These workflows are designed to assess walking, reaching, and fine motor tasks that are often affected by neurologic conditions.The research team brings deep expertise in biomechanics, rehabilitation, and artificial intelligence applications for movement analysis. Dr. Ryan Roemmich, Assistant Professor of Physical Medicine and Rehabilitation at Johns Hopkins, has more than a decade of experience in human biomechanics and motion capture. Dr. Stenum, a postdoctoral fellow, specializes in developing pose estimation workflows for clinical applications, while Dr. Celnik, Director of Physical Medicine and Rehabilitation, provides clinical oversight and patient recruitment. Together, the group is advancing AI-driven tools that aim to make motor function assessment more objective, accessible, and scalable for patients with neurologic conditions.